Metacognition is the ability to reflect on and evaluate one?s behavior. Impaired metacognition has been proposed as a major symptom for a number of psychiatric disorders such as schizophrenia, depression, generalized anxiety disorder, obsessive-compulsive disorder, and even substance abuse. Deficits in metacognition have further been found after brain lesions. Such deficits can have severe detrimental effects on patients; treating them requires insight into the computational and neural substrates of metacognition. However, efforts in this direction have been hampered by a lack of an overarching framework that can inform computational models of confidence, the measurement of metacognitive ability, and the investigation of the neural bases of metacognition. This proposal will advance a novel theoretical framework based on the concept of hierarchical noise architecture. Perceptual decision making will be used as a model system but the architecture is perfectly general and expected to apply across all domains of metacognition. According to the hierarchical noise architecture (1) sensory noise corrupts the decision-level representation of the stimulus thus affecting both the perceptual and confidence judgments, while (2) an additional metacognitive noise corrupts confidence but not the perceptual judgment. The proposed architecture makes a counter-intuitive prediction that is supported by strong preliminary data. More importantly, this architecture can be used to extract a new, model-based measure of metacognitive ability with desirable psychometric properties such as being independent from participants? bias for high or low confidence. This new measure can therefore be used to examine the effectiveness of treatments on patients? metacognitive abilities: for example, if a patient with deficiency in metacognition changes her strategy and starts using high confidence more, the new measure ? but not previous measures ? will remain unchanged. Beyond predicting new behavioral phenomena and leading to an improved measure of metacognitive ability, the hierarchical noise architecture can also elucidate the neural bases of metacognition. Specifically, the functional roles postulated by the hierarchical noise model can be linked directly to the functions of large-scale brain networks and especially the central executive and the salience networks. These links will be established using a variety of techniques such as causally interfering with different nodes of these networks, correlating metacognitive ability with the connectivity within these networks, and examining the dynamics of the inter-network communication during confidence generation. The insights gained by this proposal will have a direct link to work in the clinic by providing researchers a better tool to assess the metacognitive deficits of patients (including the effectiveness of proposed treatments) and link such dysfunction to specific brain circuits. This work is thus expected to benefit patients suffering a range of diseases from schizophrenia to depression to substance abuse.